Enable packet segment in partial redux

Reference issue

What does this implement/fix?

Additional information

Ubuntu clang version 18.1.3 -DNDEBUG -O3 -mavx2

Benchmark	Time		CPU		Iterations
test_segment_rowwise_control/1	1.12	-8.93%	1.12	-8.93%	623303252	8.73%
test_segment_rowwise_control/8	3.16	-12.03%	3.16	-12.03%	220053279	14.35%
test_segment_rowwise_control/64	72.3	-1.24%	72.2	-1.11%	9705963	1.11%
test_segment_rowwise_control/512	14123	-0.57%	14121	-0.59%	49681	0.36%
test_segment_rowwise_control/1024	245572	-0.83%	245503	-0.82%	2796	1.25%
test_segment_rowwise_1/1	1.13	-9.73%	1.13	-9.73%	624569605	9.44%
test_segment_rowwise_1/8	3.16	-11.71%	3.16	-11.71%	221901482	13.36%
test_segment_rowwise_1/64	72.3	-1.38%	72.3	-1.38%	9686812	0.65%
test_segment_rowwise_1/512	14104	-0.50%	14101	-0.48%	49521	0.64%
test_segment_rowwise_1/1024	245312	-0.70%	245256	-0.68%	2834	-0.88%
test_segment_rowwise_half/1	4.5	-42.00%	4.5	-42.00%	156829481	70.60%
test_segment_rowwise_half/8	11.8	-47.80%	11.8	-47.80%	58620389	94.46%
test_segment_rowwise_half/64	120	-22.08%	120	-22.08%	5777833	29.63%
test_segment_rowwise_half/512	9379	-4.62%	9376	-4.61%	74755	4.94%
test_segment_rowwise_half/1024	64958	-3.75%	64956	-3.77%	10817	2.12%
test_segment_rowwise_m1/1	9.64	-65.25%	9.63	-65.21%	72592615	188.59%
test_segment_rowwise_m1/8	20	-64.90%	20	-64.90%	35250256	183.58%
test_segment_rowwise_m1/64	174	-35.63%	174	-35.63%	3999856	55.68%
test_segment_rowwise_m1/512	13056	-6.36%	13055	-6.37%	53592	7.75%
test_segment_rowwise_m1/1024	76313	-2.59%	76298	-2.61%	9102	1.90%

#include <benchmark/benchmark.h>
#include <Eigen/Core>
using namespace Eigen;

using T = float;
using Mat = MatrixX<T>;
using Vec = VectorX<T>;

static void test_segment_rowwise_control(benchmark::State& state) {
  constexpr Index PacketSize = internal::packet_traits<T>::size;
  Index n = numext::round_down(state.range(0), PacketSize);
  Mat A(n, n);
  Vec b(n);
  A.setRandom();
  for (auto s : state) {
    b = A.rowwise().sum();
    benchmark::DoNotOptimize(A);
    benchmark::DoNotOptimize(b);
  }
}

static void test_segment_rowwise_1(benchmark::State& state) {
  constexpr Index PacketSize = internal::packet_traits<T>::size;
  Index n = numext::round_down(state.range(0), PacketSize);
  Mat A(n, n);
  Vec b(n);
  A.setRandom();
  for (auto s : state) {
    b = A.rowwise().sum();
    benchmark::DoNotOptimize(A);
    benchmark::DoNotOptimize(b);
  }
}

static void test_segment_rowwise_half(benchmark::State& state) {
  constexpr Index PacketSize = internal::packet_traits<T>::size;
  Index n = numext::round_down(state.range(0), PacketSize) + PacketSize / 2;
  Mat A(n, n);
  Vec b(n);
  A.setRandom();
  for (auto s : state) {
    b = A.rowwise().sum();
    benchmark::DoNotOptimize(A);
    benchmark::DoNotOptimize(b);
  }
}

static void test_segment_rowwise_m1(benchmark::State& state) {
  constexpr Index PacketSize = internal::packet_traits<T>::size;
  Index n = numext::round_down(state.range(0), PacketSize) + PacketSize - 1;
  Mat A(n, n);
  Vec b(n);
  A.setRandom();
  for (auto s : state) {
    b = A.rowwise().sum();
    benchmark::DoNotOptimize(A);
    benchmark::DoNotOptimize(b);
  }
}

BENCHMARK(test_segment_rowwise_control)->Range(1<<0, 1<<10);
BENCHMARK(test_segment_rowwise_1)->Range(1<<0, 1<<10);
BENCHMARK(test_segment_rowwise_half)->Range(1<<0, 1<<10);
BENCHMARK(test_segment_rowwise_m1)->Range(1<<0, 1<<10);
BENCHMARK_MAIN();